In this dissertation negations and modal verbs are investigated with dynamic measurements that look at human comprehension in real time. Evidence is used to argue that accumulation of activation accrues over time, at times for many things in parallel. The only way to detect differences in processing is by looking at the dynamic signal, and thus dense-sampling measurements like eyetracking and computer mousetracking are necessary to look at high-level cognition. Within a broader framework, it is argued that accumulation implies both inflow and outflow of information in parallel, and information processing may better be characterized by a metaphor of flowing water rather than the traditional box and arrow approach to characterizing mechanisms of the mind.

By using differential equations to simulate cognition at the level of abstract mental constructs, the Complex Dynamical Integrated Episode Processing Hypothesis (CDIE) sheds new light on how ordered events are encoded in episode processing. It assumes that during the episode processing current distribution of activated information is influenced by external inputs of information in a complex way due to additional influences of spatiotemporal context and spontaneous decay. In particular, it integrates perception and existing memory, attention related to time based decay modification and as possible additional simple source of information, forgetting (including both constant and non-constant time-based decay and interference), relationship of this decay and item distinctiveness, false memories resulting from three possible mechanisms and their suppression, etc. Cognition is treated as a complex dynamical system and this model for the first time in the literature integrates all of the mentioned classical concepts and sheds light on their temporal dynamics. The following are some of the unique contributions of this project.

Episode processing is currently not a particularly well identified traditional research area. Nonetheless, there exists a continuity in cognitive processing explored in many other areas and "episode processing" or the "episode processor" are defined as a complex system changing in time.

This dynamical process model is an innovative application of an infinite-dimensional dynamical system to the ordered cognitive events in episode processing. On the one hand, while dynamical systems (DS) theory has been recognized as an important tool in cognitive science, this kind of infinite-dimensional system, where a continuum of values evolves in time, has not been heavily studied by modelers or philosophers discussing dynamical approaches to non-neural level cognitive science.

The model assumes that order in time is cognitively treated as order in space. A complex systems differential equations approach to time-as-space ideas of cognitive linguistics has not been quantitatively implemented in cognitive science and it shows promise to open an entirely new area of research in cognitive linguistics.

It is often argued that the egocentric perspective is the default perspective

in natural language use. However, there are many factors that influence which

perspective a person takes. In this dissertation I look at various environmental,

biological, social, and linguistic cues that point toward another hypothesis for

consideration than just the default perspective of egocentricism. It is probable

that there is a bias toward the egocentric perspective given we are constrained to

a body which provides us with many more experiences of an egocentric

perspective but we are able to take other perspectives with relative ease and

frequency. These occurrences of an other-centric perspective suggest a graded

spatial attraction toward competing perspectives that result in a continuous,

rather than a stage-based, account of perspective taking and encourage a more

all-inclusive theoretical framework for taking a perspective.

Similarity and categorization are central phenomena in cognitive science. Despite its relevance, similarity is a poorly understood process. The current work proposes a new theory of similarity and how to unify it with categorization. Similarity is defined as the overlap of neuronal population codes that represent features of objects being compared. The experiments investigate how perceptual and conceptual information help to constrain similarity. A neural network model is designed that accurately predicts human performance on similarity judgments.

Convergence of multisensory information can improve the likelihood of detecting and responding to an event as well as more accurately identifying and localizing it. The ubiquitous nature of crossmodal processing is observed in everything from basic signal detection to speech recognition. Although we observe multisensory integration throughout various modalities, this dissertation reviews and discusses research focusing primarily on the relationship between the auditory-linguistic system and visual systems. A series of experiments and simulations presented here show graded effects of crossmodal processing that are reflected in reaction time data and motor output, measured through streaming x-y coordinates from eye-movements. A model simulates and makes predictions about real-time crossmodal processing that argue against the traditional serial and parallel approach to visual attention and supports a perspective with a single underlining mechanism. A purely parallel process is introduced as a means for reconciling both traditional and continuous accounts of visual attention. A broad philosophical discussion follows, in which an integrative and continuous approach to crossmodal processing is proposed and discussed.

Similarity and categorization are central phenomena in cognitive science. Despite its relevance, similarity is a poorly understood process. The current work proposes a new theory of similarity and how to unify it with categorization. Similarity is defined as the overlap of neuronal population codes that represent features of objects being compared. The experiments investigate how perceptual and conceptual information help to constrain similarity. A neural network model is designed that accurately predicts human performance on similarity judgments.